The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a Node.js script (scripts/relay.mjs) to execute the kimi and git command-line tools. The script uses standard child process modules and avoids shell execution, correctly implementing the skill's primary purpose of task delegation and workspace monitoring.
[EXTERNAL_DOWNLOADS]: The skill documentation refers to the official Kimi Code CLI from Moonshot AI, a well-known provider. It provides instructions for installation via Homebrew or official documentation sites.
[DATA_EXFILTRATION]: The relay script includes a security note warning users that task briefs are passed as command-line arguments and may be visible in process monitoring tools. This correctly identifies a common operational security consideration and provides appropriate mitigation advice.
[SAFE]: The skill implementation involves processing task 'briefs' which could theoretically contain malicious instructions. However, the skill explicitly mandates a multi-step review process where the agent must verify changes, re-run tests, and perform a manual commit, which effectively prevents automated exploitation through the delegated tool.

kimi-delegate